Skip to main content ホーム クリエイター mdbabumiamssm llms-universal-life-science-and-clinical-skills- bio-workflows-crispr-editing-pipeline
bio-workflows-crispr-editing-pipeline End-to-end CRISPR experiment design from target selection to delivery-ready constructs. Covers guide RNA design, off-target assessment, and specialized editing strategies including knockouts, base editing, and HDR knockins. Use when designing complete CRISPR editing experiments for gene knockout, correction, or tagging.
インストールへ移動 Skills Marketplace コミュニティが作成したAIスキルを発見・探索
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/mdbabumiamssm/LLMs-Universal-Life-Science-and-Clinical-Skills- --skill bio-workflows-crispr-editing-pipelineコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Zipをダウンロード ダウンロード中... このリポジトリの他の Skills Operate MedSAM2 for promptable segmentation of 3D medical images and medical videos, including CT lesion propagation, MRI volumes, RECIST-guided prompts, efficient CPU-oriented variants, training, and 3D Slicer integration. Use when generating or validating volumetric masks from sparse prompts or propagating masks through image slices or video frames.
Build reproducible healthcare imaging pipelines with Project MONAI for DICOM, NIfTI, pathology, and multidimensional imaging tasks including preprocessing, augmentation, training, sliding-window inference, evaluation, model bundles, labeling, and deployment. Use when implementing medical image classification, segmentation, registration, detection, generative, or foundation-model workflows in PyTorch.
Operate Google TxGemma prediction and chat models for therapeutic property prediction across small molecules, proteins, nucleic acids, diseases, targets, and cell lines. Use when formatting Therapeutics Data Commons tasks, choosing TxGemma model size or variant, running local or Model Garden inference, fine-tuning on private therapeutic data, or evaluating TxGemma in drug-discovery workflows.
mdbabumiamssm
mdbabumiamssm/LLMs-Universal-Life-Science-and-Clinical-Skills-
GitHub リポジトリを開く name bio-workflows-crispr-editing-pipeline description End-to-end CRISPR experiment design from target selection to delivery-ready constructs. Covers guide RNA design, off-target assessment, and specialized editing strategies including knockouts, base editing, and HDR knockins. Use when designing complete CRISPR editing experiments for gene knockout, correction, or tagging. tool_type mixed primary_tool crisprscan workflow true depends_on ["genome-engineering/grna-design","genome-engineering/off-target-prediction","genome-engineering/base-editing-design","genome-engineering/prime-editing-design","genome-engineering/hdr-template-design"] qc_checkpoints [{"after_grna_design":"Activity score >0.6, no poly-T runs, GC 40-70%"},{"after_offtarget":"Specificity score >0.7, no coding off-targets with <3 mismatches"},{"after_template":"Homology arms verified, PAM disrupted in donor"}]
CRISPR Editing Pipeline
Complete workflow for CRISPR experiment design: from target gene to delivery-ready constructs with branching paths for different editing strategies.
Workflow Overview
Target Gene/Position
|
v
[1. Guide RNA Design] --> CRISPRscan / Rule Set 2 / DeepCRISPR
|
v
[2. Off-Target Assessment] --> Cas-OFFinder + CFD scoring
|
v
Decision Point: What type of edit?
|
+---+-------------------+--------------------+
| | |
v v v
[3a. Knockout] [3b. Base Editing] [3c. Knockin]
Standard Cas9 CBE/ABE design HDR template
Frameshift C>T or A>G with homology arms
| | |
v v v
Final Constructs with Validation Primers
Prerequisites
pip install crisprscan biopython pandas numpy matplotlib
conda install -c bioconda primer3-py cas-offinder
pip install crisprtools
Primary Path: Gene Knockout
Step 1: Guide RNA Design
from Bio import SeqIO
from Bio.Seq import Seq
import pandas as pd
import re
def find_guides (sequence, pam='NGG' ):
'''Find all potential gRNA target sites with NGG PAM.'''
guides = []
seq_str = str (sequence).upper()
for match in re.finditer(r'(?=([ATCG]{20}[ATCG]GG))' , seq_str):
pos = match .start()
target = match .group(1 )[:20 ]
pam_seq = match .group(1 )[20 :23 ]
guides.append({
'sequence'
'pam'
'position'
'strand'
'+'
'full_target'
match
1
for
match
in
r'(?=(CC[ATCG][ATCG]{20}))'
match
match
1
str
3
23
str
0
3
'sequence'
'pam'
'position'
'strand'
'-'
'full_target'
return
def
score_guide
guide_seq
'''Score guide using Rule Set 2-like heuristics.'''
0.5
'G'
'C'
len
if
0.4
0.7
0.2
elif
0.3
or
0.8
0.2
if
'TTTT'
in
0.3
if
1
'G'
0.1
if
2
'GG'
0.1
11
20
'G'
'C'
len
if
0.4
0.7
0.1
return
min
1.0
max
0.0
'''ATGGATTTATCTGCTCTTCGCGTTGAAGAAGTACAAAATGTCATTAATGCTATGCAGAAAATCTTAGAGT
GTCCCATCTGTCTGGAGTTGATCAAGGAACCTGTCTCCACAAAGTGTGACCACATATTTTGCAAATTTTG'''
'\n'
''
'activity_score'
'sequence'
'activity_score'
0.6
'activity_score'
False
print
f'Found {len (good_guides)} high-scoring guides'
print
'sequence'
'position'
'strand'
'activity_score'
10
Step 2: Off-Target Assessment import subprocess
from pathlib import Path
def run_cas_offinder (guides_df, genome_fasta, output_dir, max_mismatches=4 ):
'''Run Cas-OFFinder for off-target detection.'''
output_dir = Path(output_dir)
output_dir.mkdir(parents=True , exist_ok=True )
input_file = output_dir / 'cas_offinder_input.txt'
with open (input_file, 'w' ) as f:
f.write(f'{genome_fasta} \n' )
f.write('NNNNNNNNNNNNNNNNNNNNNGG\n' )
for _, row in guides_df.iterrows():
f.write(f"{row['sequence' ]} NNN {max_mismatches} \n" )
output_file = output_dir / 'offtargets.txt'
subprocess.run([
'cas-offinder' , str (input_file), 'C' , str (output_file)
], check=True )
offtargets = pd.read_csv(output_file, sep='\t' , header=None ,
names=['pattern' , 'chromosome' , 'position' , 'target' ,
'strand' , 'mismatches' ])
return offtargets
def calculate_specificity_score (guide_seq, offtargets_df ):
'''Calculate CFD-based specificity score.'''
guide_offtargets = offtargets_df[offtargets_df['pattern' ].str .contains(guide_seq[:10 ])]
if len (guide_offtargets) == 0 :
return 1.0
penalty = 0
for _, ot in guide_offtargets.iterrows():
mm = ot['mismatches' ]
if mm == 0 :
penalty += 1.0
elif mm == 1 :
penalty += 0.5
elif mm == 2 :
penalty += 0.2
elif mm == 3 :
penalty += 0.1
else :
penalty += 0.05
return max (0 , 1 - penalty / 10 )
good_guides['specificity_score' ] = good_guides['sequence' ].apply(
lambda x: calculate_specificity_score(x, pd.DataFrame())
)
good_guides['combined_score' ] = (good_guides['activity_score' ] * 0.5 +
good_guides['specificity_score' ] * 0.5 )
final_guides = good_guides.sort_values('combined_score' , ascending=False ).head(5 )
Step 3a: Knockout Design (Frameshift) def design_knockout (guide_row, target_sequence ):
'''Design knockout experiment with validation primers.'''
guide_seq = guide_row['sequence' ]
position = guide_row['position' ]
cut_site = position + 17 if guide_row['strand' ] == '+' else position + 6
left_start = max (0 , cut_site - 100 )
right_end = min (len (target_sequence), cut_site + 100 )
return {
'guide_sequence' : guide_seq,
'pam' : guide_row['pam' ],
'cut_site' : cut_site,
'expected_outcome' : 'Frameshift indel' ,
'validation_amplicon_start' : left_start,
'validation_amplicon_end' : right_end
}
ko_design = design_knockout(final_guides.iloc[0 ], gene_seq.replace('\n' , '' ))
print ('Knockout Design:' )
for k, v in ko_design.items():
print (f' {k} : {v} ' )
Step 3b: Base Editing Design (CBE/ABE) def design_base_edit (target_position, target_sequence, edit_type='CBE' ):
'''Design base editing experiment.
CBE: C>T conversion (or G>A on opposite strand)
ABE: A>G conversion (or T>C on opposite strand)
Editing window: positions 4-8 in the protospacer (counting from PAM-distal)
'''
guides = find_guides(target_sequence)
suitable_guides = []
for _, guide in guides.iterrows():
guide_start = guide['position' ]
guide_end = guide_start + 20
if guide['strand' ] == '+' :
window_start = guide_start + 3
window_end = guide_start + 8
else :
window_start = guide_end - 8
window_end = guide_end - 3
if window_start <= target_position <= window_end:
target_base = target_sequence[target_position].upper()
if edit_type == 'CBE' and target_base in ['C' , 'G' ]:
suitable_guides.append(guide)
elif edit_type == 'ABE' and target_base in ['A' , 'T' ]:
suitable_guides.append(guide)
return pd.DataFrame(suitable_guides)
target_pos = 45
cbe_guides = design_base_edit(target_pos, gene_seq.replace('\n' , '' ), 'CBE' )
print (f'Found {len (cbe_guides)} CBE-compatible guides' )
Step 3c: Knockin Design (HDR Template) def design_hdr_template (guide_row, target_sequence, insert_sequence,
homology_arm_length=800 ):
'''Design HDR donor template with homology arms.
Homology arm length: 800bp is standard for plasmid donors.
For ssODN, use 30-60bp arms.
'''
cut_site = guide_row['position' ] + 17 if guide_row['strand' ] == '+' else guide_row['position' ] + 6
left_arm_start = max (0 , cut_site - homology_arm_length)
left_arm = target_sequence[left_arm_start:cut_site]
right_arm_end = min (len (target_sequence), cut_site + homology_arm_length)
right_arm = target_sequence[cut_site:right_arm_end]
guide_seq = guide_row['sequence' ]
pam_position_in_arms = cut_site - left_arm_start + 3
donor = left_arm + insert_sequence + right_arm
return {
'guide_sequence' : guide_seq,
'cut_site' : cut_site,
'left_arm' : left_arm,
'right_arm' : right_arm,
'insert' : insert_sequence,
'donor_template' : donor,
'donor_length' : len (donor),
'note' : 'Remember to mutate PAM in donor to prevent re-cutting'
}
gfp_sequence = 'ATGGTGAGCAAGGGCGAGGAG...'
hdr_design = design_hdr_template(final_guides.iloc[0 ], gene_seq.replace('\n' , '' ), 'FLAG_TAG' , 50 )
print ('HDR Design:' )
print (f" Left arm length: {len (hdr_design['left_arm' ])} " )
print (f" Right arm length: {len (hdr_design['right_arm' ])} " )
print (f" Total donor length: {hdr_design['donor_length' ]} " )
Visualization import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import numpy as np
def plot_guide_landscape (guides_df, gene_length, exon_coords=None ):
'''Visualize guide positions and scores along gene.'''
fig, axes = plt.subplots(2 , 1 , figsize=(14 , 6 ), gridspec_kw={'height_ratios' : [1 , 2 ]})
ax1 = axes[0 ]
ax1.axhline(y=0.5 , color='gray' , linewidth=10 , solid_capstyle='butt' )
if exon_coords:
for start, end in exon_coords:
ax1.axhline(y=0.5 , xmin=start/gene_length, xmax=end/gene_length,
color='steelblue' , linewidth=20 , solid_capstyle='butt' )
ax1.set_xlim(0 , gene_length)
ax1.set_ylim(0 , 1 )
ax1.set_ylabel('Gene' )
ax1.set_xticks([])
ax1.set_yticks([])
ax2 = axes[1 ]
colors = ['green' if s > 0.6 else 'orange' if s > 0.4 else 'red'
for s in guides_df['activity_score' ]]
ax2.scatter(guides_df['position' ], guides_df['activity_score' ],
c=colors, s=50 , alpha=0.7 )
ax2.axhline(y=0.6 , color='green' , linestyle='--' , alpha=0.5 , label='Threshold' )
ax2.set_xlim(0 , gene_length)
ax2.set_ylim(0 , 1 )
ax2.set_xlabel('Position (bp)' )
ax2.set_ylabel('Activity Score' )
ax2.legend()
plt.tight_layout()
plt.savefig('guide_landscape.pdf' )
return fig
plot_guide_landscape(guides, len (gene_seq.replace('\n' , '' )),
exon_coords=[(0 , 50 ), (70 , 130 )])
Parameter Recommendations Step Parameter Value Rationale Guide design Activity score >0.6 Standard threshold for reliable editing Guide design GC content 40-70% Optimal for binding and Cas9 activity Off-target Max mismatches 4 Catches most relevant off-targets Off-target Specificity score >0.7 Acceptable off-target profile Base editing Window positions 4-8 Optimal for BE3/BE4, ABE7.10 HDR Homology arms 800bp Standard for plasmid donors HDR (ssODN) Homology arms 30-60bp For single-strand oligo donors
Troubleshooting Issue Likely Cause Solution No high-scoring guides GC-poor region Expand search region, consider Cas12a Many off-targets Repetitive sequence Use high-fidelity Cas9 (eSpCas9, HiFi) Low HDR efficiency NHEJ dominant Add NHEJ inhibitors, use ssODN Base editing outside window Guide position Redesign with target in positions 4-8 Bystander edits Multiple C/A in window Design guides with single target base
Output Files File Description guides_ranked.tsvAll guides with activity and specificity scores offtargets.txtCas-OFFinder results knockout_design.jsonKO guide and validation primers base_edit_design.jsonCBE/ABE design with editing window hdr_template.fastaDonor template sequence guide_landscape.pdfVisualization of guide positions
Related Skills
genome-engineering/grna-design - Detailed scoring algorithms
genome-engineering/off-target-prediction - Cas-OFFinder and CFD
genome-engineering/base-editing-design - CBE/ABE specifics
genome-engineering/prime-editing-design - pegRNA design
genome-engineering/hdr-template-design - Donor optimization
primer-design/primer-basics - Validation primer design
